ReEscreve: A Translator-Friendly Multi-Purpose Paraphrasing Software Tool

REESCREVE
A TRANSLATOR-FRIENDLY MULTI-PURPOSE
PARAPHRASING SOFTWARE TOOL
MT Summit XII
Workshop - Beyond Translation Memories: New Tools for Translators
Anabela Barreiro & Luís Miguel Cabral Ottawa, August 29, 2009
Anabela Barreiro Luís Miguel Cabral
LINGUATECA – FCCN LINGUATECA - SINTEF ICT
1

 ReEscreve
 Why translator friendly?
 Why multi-purpose?
 Current characteristics and the future?
 Motivation for its development
 Importance of linguistic quality control
 Usefulness of style editing
 Need to automate the language control process
 Automated style editors
 Paraphrases
 Functions and relevance to NLP and (machine) translation
 Examples – word reduction
 ReEscreve: resources and interface
OUTLINE
MT Summit XII
2

WHY TRANSLATOR-FRIENDLY?
ReEscreve is intended to:
 Help translators to fine-tune the source text to their special translation needs
 Help make the source text more consistent, more fluent and in compliance
with the quality control requirements of the target language
 Help speed up and facilitate translation tasks by allowing the translator to
choose the expressions that are more appropriate for the text being translated
 Help provide alternatives and probably more useful meanings in the context of
a professional document (instead of the most obvious or literal meanings)
 Fill the gap for translators who are not true linguists, but terminology
specialists in specific fields
 Help review the target text – can be used in automated post-editing
 Be combined with and supplement translation memories and MT
MT Summit XII
3

WHY MULTI-PURPOSE?
MT Summit XII
ReEscreve can be used for various different purposes:
 Authoring-aid
 Stylistic editing
 Pre and post-editing for MT
 Controlled language
 Linguistic quality control
 Translation aid
 Pedagogical exercises
 Etc.
4

REESCREVE CURRENT CHARACTERISTICS
 It is a prototype – with an interface – a paraphraser that provides suggestions and help users to
change, simplify or clarify their texts
 It has only been tested with support verb constructions (SVC), but it is being extended to
progressively paraphrase different types of linguistic phenomena (adverbials, passives, relatives, etc.)
 At the current stage, it transforms, with a 93.4% precision, SVC into verbs or semantically
equivalent expressions
 It helps reduce ambiguity and the number of words, and eliminates unnecessary complexity
from the source text
 It works bilingually (EN-PT) for SVC – additional testing and improvements are in progress
 It still requires linguistic quality assurance – it runs on lexica that needs to be extended and
reviewed
 ReEscreve is freely accessible on the Internet at: http://www.linguateca.pt/ReEscreve/
MT Summit XII
5

 Here exemplified for the purpose of helping or simplifying machine translation -
interface, functionalities, and application to controlled writing and post-editing
However… in the future…
 Authoring and stylistic aid for word processing applications
 Linguistic quality control
 Language optimization
 Applicability to MT – being experimentally applied to a prototype MT system
proposed in Barreiro, 2008
REESCREVE IN THE FUTURE?
MT Summit XII
6

MOTIVATION TO DEVELOP REESCREVE
 Translation memories are insufficient to support translators in their translation
tasks and do not incorporate linguistic knowledge
 In the current global market, the increasing need for translation and the demand
to be simultaneously quick and rigorous requires more robust and sophisticated tools
 In future ventures, linguistic knowledge and linguistic quality control will be a
requirement for language technology tools
 Paraphrasers will play an important role in the next generation of natural
language processing tools, including translation aids
MT Summit XII
7

 Prioritize clarity and integrity of language
 Verify fluency, sentence length and lexical choices
 Allow sentence, clause, expression and word reorder
 Eliminate “wordiness”, slang, jargon, etc.
 Simplify complex and hard to understand ways of expression
 Eliminate inconsistencies
 Use style paraphrasing strategies
IMPORTANCE OF LINGUISTIC QUALITY
CONTROL AND STYLE EDITING
MT Summit XII
8

 Current tools contemplate spell checking, simple grammar checking,
ortographic and morphological variance, and synonyms
 Future tools will require the following techniques:
 Word reduction = simplification
 Linguistic information for a better understanding of the meaning of the
sentence, such as lexical realization of null subject in certain cases
 Word order (passive into active, etc.)
 Word and expression replacement with synonyms and paraphrases
AUTOMATED STYLE EDITORS
MT Summit XII
9

 Linguistic richness = many ways of “saying the same thing” using different
words, expressions or sentences
 Human beings use different linguistic strategies to express knowledge of the
real world (referential paraphrasing – related to social-cultural factors)
 Alternative expressions (≠ syntax) to express the same meaning (=
semantics)
 Retain “approximate conceptual equivalence” [Barzilay, 2003]
 Paraphrasing – technique to articulate an idea or convey information in an
alternative way, making undertanding easier and making texts more
valueable to their audience
PARAPHRASES
MT Summit XII
10

 Explanatory-referential – to define terms or concepts of a foreign country and
different social-cultural realities
 Interpretational – to paraphrase technical terms ou complex concepts/ideas into
simple words or expressions easy to understand for the ordinary language user, non-
specialist or less familiar with certain terminologies ou linguistic complexity
 Clarifying – to communicate with clarity improving understanding
 Simplification – to express ideas and concepts using simple grammar, eliminating
repetitions and anything that does not add new meaning
 Reduction – to make a text more succint, concise and less “wordy”
 Value adding – to add value/linguistic richness to a sentence or text without making it
more complex
 Style change – to establish a different style, more or less formal more or less direct,
metaphorical, consistent, uniform, etc.
 Writing practice – used as pedagogical exercise
 Translation – to translate into other language maintaining fluency in that language
MT Summit XII
PARAPHRASING FUNCTIONS
11

13
IMPACT OF PARAPHRASES IN MT
MT Summit XII
AH… THAT’S
BETTER!
MT SYSTEMS
MISHANDLING SVC
-
LITERAL TRANSLATION
OF MWE
WE PROMPTLY VERIFY THE SUPERIOR QUALITY
OF THE RESULTS ON DIFFERENT MT SYSTEMS
WHEN AN SVC IS REPLACED BY A VERB

Por falta de condições técnicas, ele foi removido para o Hospital das Clínicas, onde se fez uma amputação a nível de ombro.
FreeTranslation For lack of technical conditions, he was removed for the Hospital of the Clinics, where was done
an amputation in terms of shoulder.
WorldLingo Due to conditions techniques, it it was removed for the Hospital of the Clinics, where if the
shoulder level made an amputation.
Por falta de condições técnicas, ele foi transportado para o Hospital das Clínicas, onde os médicos amputaram o seu braço ao
nível do ombro.
FreeTranslation Due to conditions techniques, it it was carried to the Hospital of the Clinics, where the doctors had
amputated its arm to the level of the shoulder.
WorldLingo For lack of technical conditions, he was transported for the Hospital of the Clinics, where the
doctors amputated his arm level with the shoulder.
Por falta de condições técnicas, ele foi transportado para o Hospital das Clínicas, onde o seu braço foi amputado ao nível do
ombro.
FreeTranslation for lack of technical conditions, he was transported for the Hospital of the Clinics, where arm was
amputated level with the shoulder.
WorldLingo Due to conditions techniques, it it was carried to the Hospital of the Clinics, where its arm was
amputated to the level of the shoulder.
IMPROPER TREATMENT OF BIOMEDICAL SVC IN MT AND
IMPACT OF PARAPHRASING IN LINGUISTIC QUALITY CONTROL
MT Summit XII
14

EXAMPLES OF PARAPHRASES - WORD REDUCTION
MT Summit XII
Linguistic
Phenomenon
Expression or sentence Paraphrase
Adverbial à volta da órbita
around the orbit of the eye
de forma interactiva
in an interactive way
periorbital
periorbital
interactivamente
interactively
Relative
Clause
N0 que têm sido escritos
N0 that have been written
A velocidade a que se move a luz
Ŧ *The speed to which light moves
O papel que a Europa tem
The role that Europe plays/has
As dificuldades que temos
The difficulties we have
N0 que foram escritos > N0 escritos
N0 that were written > N0 written
A velocidade da luz
The speed of light
O papel da Europa
The role of Europe ≡ Europe’s role
As nossas dificuldades
Our difficulties
If clause se for necessário
if it is necessary
se necessário
if necessary
Named
Entity
A rainha de Inglaterra
The queen of England
A rainha inglesa
The British queen
Noun Phrase O heróico povo português
The heroic Portuguese people
Os heróicos portugueses
The heroic Portuguese
15

 ReEscreve
 Why translator friendly?
 Why multi-purpose?
 Current characteristics and the future?
 Motivation for its development
 Importance of linguistic quality control
 Usefulness of style editing
 Need to automate the language control process
 Automated style editors
 Paraphrases
 Functions and relevance to NLP and (machine) translation
 Examples – word reduction
 ReEscreve: resources and interface
OUTLINE
MT Summit XII
16

• The deployment of the tool was partly funded by
Linguateca, as a showcase to a wider public of what
paraphrasing resources could do, if properly
developed.
• ReEscreve’s paraphrasing resources were built on
Port4NooJ large coverage resources developed by
[Barreiro, 2008]
REESCREVE: A TOOL BUILT ON TOP OF THE
PORT4NOOJ RESOURCES
MT Summit XII
17

• Originally coming from OpenLogos MT system (English-
Portuguese language pair)
• Enhanced and converted into Portuguese-English during my
PhD research [Barreiro, 2008] – inclusion of semantic relations
and paraphrases
• Released freely through Linguateca
http://www.linguateca.pt/Repositorio/Port4Nooj/
• Documentation at
http://www.linguateca.pt/Repositorio/Port4NooJ/ResourcesOverview.pdf
PORT4NOOJ:
RESOURCES FOR PORTUGUESE PROCESSING
MT Summit XII
18

• a general dictionary of approximately 40,000 single word entries which includes:
– (optional) inflectional paradigm information
– (optional) derivational paradigm information
– syntactic-semantic information
– (optional) support verb information
– (optional) argument-structure information
– English translation for each entry
• 13,000 compounds (need to be assigned correct inflectional paradigms)
• morphological rules which cover inflection and derivation
• a grammar to process contractions)
• syntactic grammars to recognize, annotate, paraphrase and translate support verb
constructions, adverbs, relatives, passives, and possessives
• a grammar to annotate and disambiguate pronouns
• a grammar to recognize and annotate named entities
• a grammar to recognize and translate dates from Portuguese into English
MT Summit XII
19
DATA ON PORT4NOOJ
VALIDATION DISCLAIMER
• These resources still require extensive validation and evaluation
• We do not know as yet if we will get funding for it

ENHANCED DICTIONARY ENTRIES WITH SYNTACTIC-
SEMANTIC INFORMATION
adaptar,V+FLX=FALAR+Aux=1+INOP57+Subset132+EN=adapt+VSUP=fazer+DRV=NDRV00:CANÇÃO+N
Prep=de
favor,N+FLX=MAR+Npred+AB+state+EN=favor+VSUP=fazer+NPrep=a+VRB=ajudar
amputar,V+FLX=FALAR+Aux=1+OBJTRundif21+BioMed+EN=amputate+SUJ=AG+VSUP=fazer+DRV=ND
RV00:CANÇÃO+NPrep=de+OD=BP+VSTYLE=realizar+VSTYLE=efectuar+VASP=iniciar+VASP=prosseguir+
VASP=concluir
amputar,V+FLX=FALAR+Aux=1+OBJTRundif21+BioMed+EN=amputate+VSUP=fazer+DRV=NDRV00:CAN
ÇÃO+NPrep=de+OD=BP+OI=PAT+APV=sofrer+VSTYLE=realizar+VSTYLE=efectuar+VASP=iniciar+VASP=
prosseguir+VASP=concluir
rápido,A+FLX=RÁPIDO+PV+eagerType+EN=quick+DRV=AVDRV06:RAPIDAMENTE
adoçar,V+FLX=COMEÇAR+Aux=1+OBJTRundif75+Subset604+EN=sweeten+DRV=ADRV11:VERDE+VCOP
=tornar
transplantar,V+FLX=FALAR+Aux=1+RECTR26+Subset=504+BioMed+EN=transplant+SUBJ=AG+VSUP=fa
zer+DRV=NDRV79:ANO+NPrep=de+DO=BP+VSTYLE=realizar+VSTYLE=efectuar+VASP=iniciar+VASP=pr
osseguir+VASP=concluir
transplantar,V+FLX=FALAR+Aux=1+RECTR26+Subset=504+BioMed+EN=transplant+SUBJ=AG+VSUP=fa
zer+DRV=NDRV79:ANO+NPrep=de+DO=BP+IO=PAT+APV=sofrer+VSTYLE=realizar+VSTYLE=efectuar+V
ASP=iniciar+VASP=prosseguir+VASP=concluir
médico,N+FLX=ANO+AN+des+med+AG+EN=doctor
médico,N+FLX=ANO+AN+des+med+AG+EN=physician
doente,N+FLX=ANO+AN+des+med+PAT+EN=patient
20MT Summit XII

Recognition and paraphrasing of support verb constructions
(SVC/ morphossyntactic and semantically related verbs)
GRAMMAR AND CONCORDANCE TO RECOGNIZE AND
PARAPHRASE SUPPORT VERB CONSTRUCTIONS
CONTROLLED
LANGUAGE
&
TEXT EDITING
MT Summit XII
21

GRAMMAR TO RECOGNIZE AND PARAPHRASE
BIOMEDICAL SVC
22MT Summit XII

PARAPHRASING FOR LINGUISTIC QUALITY CONTROL
AND TEXT STYLISTICS
ARG0 = AGENT (AG)
Elementary SVC > Lexical Verb – fazer uma amputação = amputar
(to amputate)
Elementary SVC > non-elementary SVC - realizar/efectuar uma amputação
(to perform an amputation)
ARG0 = PATIENT (PAT)
Submeter-se/ser submetido a uma operação (to undergo surgery)
Ser operado (to be operated)
Elementary SVC > non-elementary SVC - realizar/efectuar uma operação
(to perform an operation)
23MT Summit XII

RECOGNITION AND MONOLINGUAL PARAPHRASING
OF BIOMEDICAL-RELATED SVC
24MT Summit XII

ReEscreve online
Combine
Text and
Suggestions
Input Text
Noojapply
Paraphrase
Suggestions Port4NooJ
Dictionaries Grammars
REESCREVE ARCHITECTURE
MT Summit XII
25

Let’s try this text!
REESCREVE – INTERFACE
MT Summit XII
Interactive use of ReEscreve
ReEscreve can be used in word processing
applications or linguistic quality control for
both source and target texts
26

REESCREVE SUGGESTIONS
FOR YOUR TEXT
Paraphrases of SVC presented by ReEscreve’s paraphrasing system
MT Summit XII
27

REWRITING YOUR OWN SUGGESTIONS
User can suggest
new expressions!
MT Summit XII
28

Log
•Original sentence : Dizia que as escadas eram uma maçada, que o senhorio estava a fazer planos para
instalar um elevador, o que seria uma vantagem quando Frances fosse lá ficar com o bebé
•Selected: Dizia que as escadas eram uma maçada, que o senhorio estava a planear instalar um
elevador, o que seria uma vantagem quando Frances fosse lá ficar com o bebé
REGISTERING USER CHOICES AND
IMPROVING THE RESOURCE
MT Summit XII
•Original sentence : Ultimamente comecei a dar ainda mais valor à privacidade e ao anonimato
desta casa.
•Sugestion: Ultimamente comecei a apreciar à privacidade e ao anonimato desta casa.
29

A REWRITTEN TEXT
Text rewritten based on the
user’s preferences
MT Summit XII
30

Support
Verb
Recognition of SVC
Precision
Recognition of SVC
Recall
Paraphrasing of SVC
Precision
Pôr 73/73 - 100% 73/100 – 73% 72/73 - 98.6%
Tomar 75/75 - 100% 75/100 – 75% 68/73 - 93.1%
Ter 65/65 - 100% 65/100 – 65% 59/65 - 90.7%
Dar 57/60 - 95% 57/100 – 57% 46/51 - 90.1%
Fazer 43/45 – 95.5% 43/100 – 43% 40/45 - 88.8%
98.4% 62.6% 93.4%
Evaluation of
recognition and
paraphrasing
of SVC with
ReEscreve
5 support verbs
500 sentences – manually annotated
100 sentences for each support verb
PRELIMINARY EVALUATION
31MT Summit XII

Recognition of
Portuguese SVC
and translation
into English verbs
MT Summit XII
PARAMT – A PARAPHRASER FOR MT
32
MACHINE
TRANSLATION

MT Summit XII
MT WITH PARAPHRASES
33

34
 Enlargement and improvement of Port4NooJ’s semantic relations
• Assignment of tematic roles and semantic relations to named entities
Exs: médico+AG+EN=doctor
doente+PAT+EN=patient
• Enlargement of the syntactic-semantic relations between predicates and other
elements semantically related to them
Exs: operar – operação – operado
urgente – urgentemente
• Inserting synonyms, style variants, aspectual variants and enlargement of the links
between verbs and support verb constructions
 Many linguistic phenomena need to be formalized – one phenomenon at a time,
makes the linguistic knowledge database to grow!
HOW TO IMPROVE REESCREVE?
MT Summit XII

?
35
HOW TO INTEGRATE REESCREVE
IN OTHER TOOLS?
MT Summit XII

RELEVANT PUBLICATIONS
• Barreiro, A. 2008b. Make it simple with paraphrases: Automated paraphrasing for authoring aids and machine
translation. PhD dissertation. FLUP. Porto, Portugal. December 2008.
• Anabela Barreiro. "ParaMT: a Paraphraser for Machine Translation". In António Teixeira, Vera Lúcia Strube de Lima,
Luís Caldas de Oliveira & Paulo Quaresma (eds.), Computational Processing of the Portuguese Language, 8th
International Conference, Proceedings (PROPOR 2008) Vol. 5190, (Aveiro, Portugal, 8-10 de Setembro de 2008),
Springer Verlag, pp. 202-211.
• Anabela Barreiro (forthcoming). "Paraphrasing Biomedical Support Verb Constructions for Machine Translation".
In Proceedings of the 2009 International NooJ Conference (Tozeur, Tunisia, June 8-10, 2009), Cambridge Scholars
Publishing.
• Anabela Barreiro (forthcoming). "Linguistic resources and applications for Portuguese processing and machine
translation". In Proceedings of the 2008 International NooJ Conference (Budapest, Hungary, June 8-10, 2008),
Cambridge Scholars Publishing.
• Anabela Barreiro. "Port4NooJ: Portuguese Linguistic Module and Bilingual Resources for Machine Translation". In
Proceedings of the 2007 International NooJ Conference (Barcelona, Spain, June 7-9, 2007), Cambridge Scholars
Publishing.
• Anabela Barreiro & Elisabete Ranchhod. "Machine Translation Challenges for Portuguese". Linguisticæ
Investigationes 28.1 (2005), pp. 3-18. (Machine Translation, Controlled Languages and Specialised Languages).
Amsterdam/Philadelphia: John Benjamins Publishing Company. ISSN: 0378-4169.
• Anabela Barreiro. "Port4NooJ Linguistic Resources Overview". 2008.
http://www.linguateca.pt/Repositorio/Port4NooJ/ResourcesOverview.pdf
36MT Summit XII

Project ReEscreve is partially supported by Linguateca, jointly
funded by the Portuguese Government, the European Union
(FEDER and FSE), under contract ref. POSC/339/1.3/C/NAC, and
by UMIC and FCCN. The first author was supported by PhD grant
SFRH/BD/14076/2003.
ACKNOWLEDGMENTS
MT Summit XII
37

REESCREVE
A TRANSLATOR-FRIENDLY MULTI-PURPOSE
PARAPHRASING SOFTWARE TOOL
MT Summit XII
Workshop - Beyond Translation Memories: New Tools for Translators
Anabela Barreiro Luís Miguel Cabral
LINGUATECA – FCCN LINGUATECA - SINTEF ICT
38

ReEscreve: A Translator-Friendly Multi-Purpose Paraphrasing Software Tool

More Related Content

More from INESC-ID (Spoken Language Systems Laboratory - L2F) (20)

Recently uploaded (20)

ReEscreve: A Translator-Friendly Multi-Purpose Paraphrasing Software Tool